A major question in understanding the variation in DNA sequences within and between species is how important are selective constraints in determining this variation. Differential survivorship and fertility of individuals carrying different gene forms is of interest both for an understanding of the evolution of genes and from the standpoint of health, since so many human diseases are influenced, if not directly caused, by genetic variants. Except for a few cases of very strong physiological effect, it is not possible to detect selective differences by direct observation, so it has become necessary to use statistical tests on observed DNA sequence variation in populations. These tests, however, depend on a large array of assumptions that are not met in real populations. This research project will use large-scale computer simulation of populations to ask under what conditions the statistical tests presently used can actually detect selection and under what conditions an apparent detection of natural selection is really a consequence of the fact that populations deviate in various ways from the underlying assumptions of the tests about population history and structure. The ultimate goal is to prescribe the most powerful test procedures that are not unduly sensitive to unmet assumptions. [unreadable] [unreadable]